signal detection methods in argus and aers … detectio… · signal detection methods in argus and...
Post on 18-Aug-2018
227 Views
Preview:
TRANSCRIPT
Dipti Kadam DBMS Consulting
13 October 2010
Argus and Oracle AERS Focus Group
Session 24
Signal Detection Methods in Argus and AERS summarized and
compared
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Agenda
Introduction Background: Signal Detection Background: Signal Analysis Common Signal Detection Method Signal detection in Oracle AERS 4.6 Signal detection in Oracle Argus
• Overview of ARGUS Perceptive • Overview of ARGUS Insight
Signal detection analysis in external tools Future Improvements
2 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Introduction
Demand for improved public health standards
Regulatory bodies monitor drugs very closely
Product development and pre-marketing review process - to define a
risk-benefit decision for product approval and clear communication
Post Marketing Surveillance for identification and evaluation of
potential AEs information in a timely and efficient manner
Ensures an accurate and integral safety profile of drugs.
Pharmaceutical companies conduct Signal detection and analysis.
3 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Background: Signal Detection
A process for identifying trends and correlations between Adverse
Events with particular drugs compared to what would be expected and compared to that for other similar drugs.
The concept is to monitor AEs which occur coincidentally and
those which are actually related to the treatment or medication being provided.
WHO defines a pharmacovigilance signal as “reported information
on a possible causal association between an adverse event and a drug, the relationship being unclear or incompletely documented previously”.
4 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Background: Signal Detection (2)
5 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
The typical process flow for Signal detection
Consultation Potential Signals
Action
Statistical output
Patient exposure
Refined Signal(S)
Background incidence
Comparative Data
Periodic Event
Listings
Identify Potential Signals
Consult Marketing
Consult SAS Programmer
Consult Database
Integrate Information
Consult Literature
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Background: Signal Analysis
To determine whether or not a Signal is valid, a threshold must be defined. This is a mathematical formula to determine whether or not an AE should be counted as part of a Signal.
Confirmed Signal may require subsequent action as appropriate.
As Signals are discovered, they must be analyzed to determine whether it represents a potential safety risk. This analysis can include: Causality Frequency Specificity vs. Sensitivity Determination of health consequences (individual outcome and
societal costs)
6 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Background: Signal Analysis
There are many types of statistical models which are used in Signal Detection analysis, but three of the most common are:
Due to the number of Adverse Event reports and the volume of the comparison reference data, some analytical method is required. Manual comparison is not feasible.
7 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
•PRR is used by Eudravigilance System
Proportional Reporting Ratio [PRR]
•BCPNN is used by the WHO
Bayesian Confidence Propagation Neural Network [BCPNN]
•MGPS is used by the FDA
Multi-item Gamma Poisson Shrinker [MGPS]
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Common Signal Detection Method in AERS and Argus
PRR (Proportional Reporting Ratio)
This is automated detection method which involves the calculation of a
very simple disproportionality method
Adverse Event: A [defined by code or groups of codes within MedDRA (ΩA)]
Treatment: T[defined by code or groups of codes within medicinal products dictionary (ΩT)]
ICSR : i, so AE on ICSR: Ai and Treatments: Ti
X (A,T)= |i: A ∩ Ai≠ Ф and T ∩ Ti ≠Ф|
So, Number of ICSRs containing an element of A and of T
PRR(A,T) = [X(A,T)/X(ΩA, T)]/[X(A, ΩT)/X(ΩA, ΩT)]*
The ordering of pairs (A, T) based on the PRR values gives a baseline against which to compare other methods of signal detection (i.e. other ordering of (A,T)).
8 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010 Current State of Signal Detection in AERS 4.6
AERS 4.6 Built in functionality for Signal detection and reporting
Built in Folder Queries Under Pharmacovigilance Folders.
Built in reports such as Signal Identification report
According to AERS 4.6.1 Users Guide, pg 9-4, “the Signal Identification Report is designed to compare safety profiles between date periods to help identify potential changes in a product's safety profile.”
Implication that there is not a comparative method against other external data sources.
According to the AERS 4.6.1 Release Content Document, pg 15, “AERS 4.6 has been integrated with the Druglogic’s Qscan. You are now able to pass a set of cases into Qscan for further analysis and visualization, then identify and review the cases of interest within AERS.”
9 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Possible Integration Strategies for Oracle BI and AERS
Oracle BI has ability to scale and process large amounts of data through daily processing cycles, called ETLs (Extract Transform-Load)
The ETL can perform a direct extraction from an Oracle AERS instance, using the same queries which drive the Signal Identification Report
Additionally, external source data can be loaded from XML files from FDA, WHO or EU
Mapping the analysis method directly into the Oracle BI (Siebel Analytics) engine can be done for Signal detection analysis
BI Dashboard can be developed with the actual threshold parameters as well as specificity and sensitivity percentages
Each analysis can be stored and replayed in scenarios. This makes longer-term signal detection possible with very little efforts.
10 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Signal Detection with Oracle Argus
11 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Integration with
ARGUS SAFETY and ARGUS Insight
Oracle ARGUS
Perceptive
Oracle Argus Perceptive when integrated with Argus Insight, builds on a comprehensive safety knowledge base can provide a proactive approach to risk management.
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Oracle Argus Perceptive
Key Features:
Dynamic Workflow
Flexible enough to fit any risk management business process
Assessor’s workflow separate from case processing
Enhanced algorithms- more choice for alerting
PRR, Chi square, Case Scoring
Work lists
Easier differentiation of active/archive & watchlisted alert
Automated Signal Detection
Analytical tools for signaling, hypothesis testing & strengthening.
12 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Ref: Presentation of Argus Perceptive by Oracle
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Oracle Argus Perceptive
Key Features (Contd.)
Searchable knowledgebase & repository
Multiple levels of escalation & notifications
Management monitoring tools & reports
Insight Linkage :Passing the event level details to Insight in preparation for more event driven reporting & querying
Primary / Secondary separation
Monitored / Ad hoc vs. mined
Linked alerts
Support for Signal Workup Document- regulatory reporting
13 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Ref: Presentation of Argus Perceptive by Oracle
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Oracle Argus Perceptive
Primary alerts based on-
Designated Medical Events (DME)
Product Important Event (PIE)
Always Expeditable Events (EXP)
Serious Unlisted Events (SUL)
User Defined Alert type
The Role of Secondary Alerts
• Directed Research and Hypothesis Generation
• Data Mining
• Generate a Case Series to process in Argus Perceptive
• Generate Output (reports, Charts) for signal workups
14 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Two Types of Alerts….
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Oracle Argus Insight
Key Features:
Query Capabilities (QBE, Filters, Advance queries, Query libraries)
Case Series
Reports output:
Standard Reports
Report Writer
Dashboard Indicators
Cubes
15 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Ref: Presentation of Argus Insight by Oracle
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Signal Detection: SAS with AERS/Argus
SAS Signal detection framework can be used as component to generate Safety Signals using data migration from AERS or Argus database to SAS Platform
This can be further analyzed using the SAS BI Suite.
16 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Signal Detection: SAS with AERS/Argus
Source Data from AERS, ARGUS, Regulatory can be extracted using SAS ETL.
Processing data can be further cleaned, organized in the staging area.
Data will be stored in the Data Warehouse.
By applying transformation rules incremental data from data warehouse can be populated in customized outbound area.
Signal definition criteria (defined threshold) can be applied
17 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Signal Detection: SAS with AERS/Argus
Contd ..
Signal can be detected on basis of Signal criteria defined by users
Analysis of signals generated by system can be done in form of report output, trend charts etc.
18 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Ref: Published paper AD06
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
19 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Future Suggestions
AERS:
Direct integration between AERS and Oracle BI is possible if a pre-loaded ETL is made available. This is relatively straightforward since both the analysis methods and the current reporting queries are known.
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Disclaimer
Information disseminated in this presentation should in no way be construed as a recommendation to use or not use any particular software or hardware product and is presented solely for informational purposes.
20 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Acknowledgements
Many thanks to the OHSUG for this opportunity to present for the OHSUG Dipti Kadam Focus Group.
Special thanks to the OHSUG Planning and Review Committee and OHSUG Dipti Kadam Focus Group Chairs for their infinite patience in receiving and expeditious review of this presentation
Thank you everyone who participated in the development of presentation.
21 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Bio:
Dipti is a Functional Consultant for DBMS Consulting in Drug Safety, Pharmacovigilance with 5+ years of experience as Clinical research professional. Prior to joining DBMS Consulting, Dipti was a Senior Executive, Pharmacovilance at SIRO Clinpharm Pvt. Ltd. and Senior Associate, Safety Data Review at Pfizer Pharmaceuticals.
22 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
23 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
Questions?
Dipti Kadam
dipti.kadam@clinicalserver.com
+46-736-433-210
+1-888-737-8818
OHSUG 2010 San Antonio, Texas: FG; Title October 2010
Thank you!!
24 ©2010 DBMS Consulting, Inc. Unauthorized Duplication is Strictly Prohibited Presented by: Dipti Kadam
top related